This action might not be possible to undo. Are you sure you want to continue?
Test equality of means The procedure known as the Analysis of Variance or ANOVA is used to test hypotheses concerning means when we have several populations. The Analysis of Variance (ANOVA) The ANOVA procedure is one of the most ANOVA is a general technique that can powerful statistical techniques be used to test the hypothesis that the means among two or more groups are equal, under the assumption that the sampled populations are normally distributed. A couple of questions come immediately to mind: what means? and why analyze variances in order to derive conclusions about the means? Both questions will be answered as we delve further into the subject. To begin, let us study the effect of temperature on a passive component such as a resistor. We select three different temperatures and observe their effect on the resistors. This experiment can be conducted by measuring all the
Introduction to ANOVA
which are the differences between before and after being subjected to the temperatures. Levels may be quantitative numbers or. in many cases. • What is a factor? The 1-way ANOVA The 2-way or 3-way ANOVA Hypotheses that can be tested in an ANOVA The alternative hypothesis is: the means . we may be interested in a third factor. 24 hours and analyze the responses. The null hypothesis is: there is no difference in the population means of the different levels of factor A (the only factor). First consider the one-way ANOVA. The temperature is called a factor. temperature and oven position. Then we measure the resistors again after. A factor is an independent treatment variable whose settings (values) are controlled and varied by the experimenter. Here we speak of a two-way or two-factor ANOVA. In this case there would be two factors. and the analysis of variance that we will be using to analyze the effect of temperature is called a one-way or one-factor ANOVA. Furthermore. say.participating resistors before placing n resistors each in three different ovens. In each of these ANOVA's we test a variety of hypotheses of equality of means (or average responses when the factors are varied). Each oven is heated to a selected temperature. The different temperature settings are called levels. there is only one factor. simply "present" or "not present" ("0" or "1"). the effect of time. The intensity setting of a factor is the level. Now we deal with a threeway or three-factorANOVA. In this example there are three levels or settings of the factor Temperature. We could have opted to also study the effect of positions in the oven. temperature. In the experiment above.
There is no difference in the means of factor A 2. and BC. The last term is for then-factor interaction and is always 1. we will discuss only the 1-way and 2-way ANOVA. For the 3-way ANOVA: The main effects are factors A.are not the same. and is always 1. There is no difference in means of factor B 3. The third term is for the number of 2-factor interactions. In what follows. . and the alternative hypothesis is the means are not equal. For the 2-way ANOVA. and so on. the number of main effects and interactions can be found by the following expression: The n-way ANOVA The first term is for the overall mean.3. AC.4. For each of the seven cases the null hypothesis is the same: there is no difference in means. the possible null hypotheses are: 1. The second term is for the number of main effects. The alternative hypothesis for case 3 is: there is an interaction between A and B. There is no interaction between factors A and B The alternative hypothesis for cases 1 and 2 is: the means are not equal. 7.1. B and C. In general. The 2factor interactions are: AB. 1-Way ANOVA overview Overview and principles This section gives an overview of the one-way ANOVA. There is also a three-factor interaction: ABC.
The variance. called the "sum of squares of treatments" and "sum of squares of error". denotes the "grand" or "overall" mean. Each ni is the number of observations for treatment i. after some algebra. Algebraically.. The variance of n measurements is given by Variance of n measurements Sums of squares and degrees of freedom where is the mean of the n measurements. and this is usually abbreviated by Total SS or SS(Total). Another name for the numerator is the "corrected sum of squares". respectively. or Total SS where k is the number of treatments and the bar over the y. The goal in this procedure is to split the total variation in the data into a portion due to random error and portions due to changes in the values of the independent variable(s). this is expressed by The guiding principle behind ANOVA is the decomposition of the sums of squares. In an analysis of variance the variation in the response measurements is partitoned into components that correspond to different sources of variation. The SS in a 1-way ANOVA can be split into two components. can be rewritten as: The first term in the numerator is called the "raw sum of squares" and the second term is called the "correction term for the mean". and the denominator is called the degrees of freedom. The numerator part is called the sum of squares of deviations from the mean.Partition response into components First we explain the principles involved in the 1-way ANOVA. The total . abbreviated as SST and SSE.
The double subscript stems from the way the data are arranged in the data table. the lengths of the rows. Random effects If the k levels of treatment are chosen at random. Whether the levels are fixed or random depends on how these levels are chosen in a given experiment. the model model equation remains the same. The definition is: A treatment is a specific combination of factor levels whose effect is to be compared with other treatments. The table is usually a rectangular array with k columns and each column consists of ni rows (however. .ni) on the i-th treatment (i = 1.3. respectively. one fortreatments and the second for error. 7. The 1-way ANOVA model and assumptions A model that describes the relationship between the response and the treatment (between the dependent and independent variables) Fixed effects model The mathematical model that describes the relationship between the response and treatment for the one-way ANOVA is given by where Yij represents the j-th observation (j = 1.. However. 2. The total SS can be written single or double subscripted. now the i's are random variables assumed to be NID(0.. It is also assumed that is chosen so that holds. 2.3.. ). The ANOVA table and tests of hypotheses about means Sums of Squares help us compute the variance estimates displayed in ANOVA Tables The sums of squares SST and SSE previously computed for the one-way ANOVA are used to form two mean squares.number of observations isN (the sum of the ni).. Y23 represents the third observation using level 2 of the factor. or the ni. is always a fixed parameter and are considered to be fixed parameters if the levels of the treatment are fixed. known as .4. Definition of "Treatment" We introduced the concept of treatment. So.2.. These mean squares are denoted by MST and MSE. i represents the i-th treatment effect and ijrepresents the random error present in the j-th observation on theith treatment. with mean zero and variance . Note on subscripting 7. These are typically displayed in a tabular form. is the common effect for the whole experiment. This is the random effects model.3.4. may be unequal). Don't be alarmed by the double subscripting. and not a random sample from a population of possible levels. The errors ij are assumed to be normally and independently (NID) distributed. . k levels). This is the fixed effects model.
we have an experiment in which each of three treatments was replicated 5 times. the degrees of freedom for treatment. The corresponding mean squares are: MST = SST / DFT MSE = SSE / DFE The test statistic. based on the chosen level and the degrees of freedom DFT and DFE.3 6.5 . In the language of Design of Experiments. Level 1 6. The critical value is the tabular value of the F distribution. to obtain mean squares Let N = ni. used in testing the equality of treatment means is: F = MST / MSE. DFE = N . the two MSE mean squares estimate the same quantity (error variance).4 Level 2 8. their ratio should be close to 1. respectively. The sample size of each group was 5. MST should be larger than MSE. DFT = k . Divide sum of squares The mean squares are formed by dividing the sum of squares by degrees of freedom by the associated degrees of freedom.9 5. In other words.1. as follows: Source SS DF MS F Treatments Error Total (corrected) SST SSE SS k-1 SST / (k-1) MST/MSE N-k SSE / (N-k) N-1 The F-test ANOVA table The word "source" stands for source of variation.k. Then. Ratio of MST and When the null hypothesis of equal means is true. Some authors prefer to use "between" and "within" instead of "treatments" and "error". The calculations are displayed in an ANOVA table.0 10. The ANOVA table also shows the statistics used to test hypotheses about the population means. If the null hypothesis is false. and the degrees of freedom for error.8 Level 3 8. and should be of approximately equal magnitude.an ANOVA Table. ANOVA Table Example A numerical example The data below resulted from measuring the difference in resistance resulting from subjecting identical resistors to three different temperatures for a period of 24 hours.
2 6. 1-Way ANOVA calculations Formulas for 1way ANOVA hand calculations Although computer programs that do ANOVA calculations now are common.041 1 Factor Interpretation of the The test statistic is the F value of 9.8 4.8 9. and by how much.1 6. Since the test statistic is much larger than the critical value.34 7.3.3 8.05. What we do not know at this point is whether the three means are all different or which of the three means is different from the other two.59. The pvalue for 9.454 F 9. so the test statistic is significant at that level.949 1. we reject the null hypothesis of equal population means and conclude that there is a (statistically) significant difference among the population means.4. the goal is to produce two variances (of treatments and error) and their ratio. the correction for the mean.9 9.59 Total (corrected) 45. • 7. for reference purposes this page describes how to calculate the various entries in an ANOVA table.4.72 8. There are several techniques we might use to further analyze the differences. 2.56 Example ANOVA table The resulting ANOVA table is Source SS DF Treatments Error 27. ANOVA table we have that F.897 17.349 14 Correction 779.452 2 12 MS 13. The various computational formulas will be shown and applied to the data from the previous example.6 4. Step 1: compute STEP 1 Compute CM. Techniques for further The populations here are resistor readings while operating analysis under the three different temperatures. CM .5.05.5 7.00325.59 is . estimating combinations of factor levels with confidence bounds multiple comparisons of combinations of factor levels tested simultaneously. 12 = 3.0 means 5. Remember. Using an of .89 (see the F distribution table in Chapter 1). These are: • • constructing confidence intervals around the difference of two means.
Hence.27.9) + (5..8 Then Step 4: compute STEP 4 Compute SSE.. the error sum of squares. SST First we compute the total (sum) for each treatment. MSE and their ratio.. SSE Here we utilize the property that the treatment sum of squares plus the error sum of squares equals the total sum of squares.897 = 17.3) + (6.5) + . + (6. + (4. total SS The total SS = sum of squares of all observations .349 ..6 T1 = (8. MSE is the mean square of . the treatment sum of squares.0) = 26.8) + ..5) = 38.7 T2 = (8. MST.4) + .3) = 42. + (9. and F MST is the mean square of treatments. MSE. SSE = SS Total .0) + (10. T1 = (6.390 SS is called the "raw" or "uncorrected " sum of squares.CM The 829..SST = 45.45. Step 3: compute STEP 3 Compute SST. F.Step 2: compute STEP 2 Compute the total SS. Step 5: Compute STEP 5 Compute MST.
5.025.error (MSE is also frequently denoted by ). These numbers are the quantities that are assembled in the ANOVA table that was shown previously.179 2.787 to 3.56 • 1 = 5.2 is: from -1.616.452/ 12 = 1. The formula for a (1. MST = SST / (k-1) = 27. .12 = 2.59 That is it. A 95% confidence interval for 3 .949 MSE = SSE / (N-k) = 17.1 is: from -0.5. Finally.247 to 5.763) or Substituting these values yields (8.34) 3.007.897 / 2 = 13.454 where N is the total number of observations and k is the number of treatments. Additional 95% confidence intervals Contrasts discussed later That is.4. Computation of the For the example.34 • • t. including how to put confidence bounds around contrasts.179(0.56 . we have the following quantities for the confidence interval formula: for 3 . Confidence intervals for the difference of treatment means Confidence intervals for the difference between two means Formula for the confidence interval This page shows how to construct a confidence interval around ( i. the confidence interval is from 1.3. A 95% confidence interval for 2 .467. compute F as F = MST / MSE = 9.604 to 4. 7.1 • 3 = 8.22 1.j) for the one-way ANOVA by continuing the exampleshown on a previous page.) 100% confidence interval for the difference between two treatment means is: where = MSE. Later on the topic of estimating more general linear combinations of means (primarily contrasts) will be discussed.836.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.